Table 4 Identifying the main taxonomies and algorithms used in AR for autonomous transportation based on relevant studies to our second research question
From: A review on action recognition for accident detection in smart city transportation systems
Authors | Model | Architecture | Model Features | ACC/AUC/DR/IOU | RMSE/MAP/MAPE | Precision | F1score | Recall | Model Comparison | ACC/AUC/DR/IOU | MAPE/MAE | Citation |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Yao et al. [86] | Future object localization | Two stream RNN Ego motion RNN | Object Localization Object detection | 73 | – | – | – | – | ConvAE | 64.3 | – | 16 |
ConvLSTMAE | 67.5 | |||||||||||
AnoPred | 64.8 | |||||||||||
Yu et al. [23] | Deep spatiotemporal graph convolutional network (DSTGCN) | Graph convolution network | Weather, road network | – | 0.34 | 82 | 85 | 89 | SVM | 0.79 | – | 25 |
SdAE | 0.81 | |||||||||||
TARPML | 0.83 | |||||||||||
Wang et al. [80] | Spatial temporal graph neural network | Spatial-GNN + GRU + Transformer | Traffic Speed Spatiotemporal Dependencies | – | 3.99 | – | – | – | FC-LSTM | – | 3.44 | 82 |
TGNN | 2.62 | |||||||||||
Bao et al. [88] | Spatiotemporal GCN (graph convolution network) | Graph convolution + RNN | Accident relevant cues | – | 53.7 | – | – | – | adaLEA | – | 52.3 | 15 |
DSA | – | 48.1 | ||||||||||
Reddy et al. [26] | Deep Q-Learning | Deep Q-Learning YOLOv3 | Car speed Distance and position | 90 | – | 87 | – | – | Yolo 3 | 85 | – | – |
Fernandez et al. [84] | Two-stream network | Two-stream convolutional networks Spatiotemporal Multiplier networks | Lane change | 90.3 | – | – | – | – | Disjoint two-stream convolution | 89.6 | – | 7 |
Ali et al. [22] | Dynamic deep spatiotemporal neural network (DHSTNet) | Graph convolution network + LSTM | Weather condition Traffic flow | – | 11.08 | – | – | – | DHSTNet Aatt–DHSTNet | – | 12.80 13.72 | 4 |
Wang et al. [21] | Spatial–temporal Mixed attention graph-based convolution model (STMAG) | GRU + mixed attention mechanism | Object detection Lane marking | – | 3.23 | – | – | – | XGBOOST | – | 3.71 | 7 |
SVR | 3.99 | |||||||||||
LSTM | 3.43 | |||||||||||
Huang et al. [25] | CNN-traffic incident management (TIM) | CNN | Car speed Road occupancy | 80 | – | – | 78 | – | RF | 76 | – | 31 |
Bortnikov et al. [92] | CNN | 3D convolutional neural network (CNN) | Optical flow Vehicle trajectory | – | – | – | 71 | – | – | – | – | 13 |
Gupta et al. [93] | Time-distributed RNN | LSTM | Temporal features Hierarchical features | 94 | – | 95 | 84 | 75 | – | – | – | 1 |
Yang et al. [94] | Feature-fused SSD detector | Single-shot multibox detector (SSD) | Detection box | – | 70.5 | – | – | – | SSD TPN | – | – | 2 |
Ijjina et al. [83] | Mask R-CNN | Deep CNN | Car speed Vehicle trajectory | DR – 71 | – | – | – | – | Deep spatiotemporal network | DR–77 | – | 24 |
You et al. [85] | Single-stream temporal action proposals (SST) | Temporal segment networks (TSNs) | - | IOU – 42.07 | – | – | – | – | R–C3D MS–TCN | – | – | 10 |
Srinivasan et al. [24] | Detection transformers and random forest classifier (DETR) | DETR + CNN | Object detection | 78.7 | – | 77 | 77 | 78 | ARRS CVABTS | DR– 50 DR– 71 | – | 1 |
Hui et al. [95] | Gaussian mixture model (GMM) | - | Vehicle detection Object tracking | – | – | – | – | – | – | – | – | 39 |
Min et al. [87] | Sparse topic model | Scale-invariant feature transform (SIFT) flow | Motion Pattern | AUC – 91.2 | – | – | – | – | GPR | 85.5 | – | 5 |
JSM | 80.2 | |||||||||||
BiLSTM | 88.1 | |||||||||||
Vatti et al. [96] | Accident detection and communication system | – | Motion pattern | – | – | – | – | – | – | – | – | 16 |